To assess doctor-patient communication in patients with glaucoma. Observational cohort study. Twenty-three ophthalmologists and 50 patients with glaucoma. Doctor-patient encounters were audio- and videotaped and analyzed using validated sociolinguistic approaches. After the visit, the doctor and the patient completed questionnaires, and patients were interviewed using a semistructured, patient-centered protocol. Summary statistics about doctor-patient encounters, assessment of alignment of attitudes between patients and doctors, and patient admission to missing doses. Physicians spent an average of 8.0 (standard deviation [SD], 3.1; median, 7.8) minutes in the room with the patient and an average of 5.8 (SD, 2.4; median, 7.5) minutes talking with the patient, delivering 70% of all spoken words and asking two thirds of all questions. Glaucoma-related discussion occupied 50% of talk time and was focused primarily on examinations and treatment (25%). One third of discussions addressed ocular issues other than glaucoma. Virtually all physician questions (94%) were closed ended. Most patient questions were about intraocular pressure (20% of visits), details of the medication regimen (20%), disease status (14%), and testing (12%). Although physicians and patients were aligned in believing that the physician should control the visit agenda, physicians tended to support greater physician control of decision making than did patients. Physicians failed to identify most patients who admitted to missing doses, a surrogate for nonadherence, stating that 10 of 13 in this category were taking drops "all" or "most" of the time. Physician interviews detected 3 of the 11 patients whose postvisit questionnaire indicated missing a dose in the last week compared with 11 of the 11 detected by the postvisit research interview. Patients who stated they had missed doses recently reported being less satisfied with the doctor-patient encounter than those who did not. Doctor-patient dialogue was universally physician centered; physicians spoke 70% of the words and asked closed-ended questions that restricted the patient's contribution to "yes/no" or brief responses. A minority of physicians ever asked patients if they had questions. In contrast with the patient-centered research interview, doctors' physician-centered communication failed to identify most patients who had missed doses.